Data-Efficient Multi-Agent Spatial Planning with LLMs

TL;DR

This paper demonstrates that large language models can be effectively used for multi-agent spatial planning tasks like taxi routing, achieving strong zero-shot performance and requiring significantly fewer environmental interactions through strategic prompting and limited fine-tuning.

Contribution

The study shows how to leverage pretrained LLMs for multi-agent decision making, introducing prompting strategies and a look-ahead algorithm that outperform existing methods with fewer interactions.

Findings

Zero-shot performance is strong with proper prompting.

Limited fine-tuning and look-ahead algorithms outperform existing approaches.

Including easy-to-compute information in prompts improves results.

Abstract

In this project, our goal is to determine how to leverage the world-knowledge of pretrained large language models for efficient and robust learning in multiagent decision making. We examine this in a taxi routing and assignment problem where agents must decide how to best pick up passengers in order to minimize overall waiting time. While this problem is situated on a graphical road network, we show that with the proper prompting zero-shot performance is quite strong on this task. Furthermore, with limited fine-tuning along with the one-at-a-time rollout algorithm for look ahead, LLMs can out-compete existing approaches with 50 times fewer environmental interactions. We also explore the benefits of various linguistic prompting approaches and show that including certain easy-to-compute information in the prompt significantly improves performance. Finally, we highlight the LLM's built-in semantic understanding, showing its ability to adapt to environmental factors through simple prompts.